Clef-Flash
Clef-Flash is a 9B multimodal model that maps a state and a schema of typed questions to decisions. It returns a probability for each allowed option, without free-form text generation or output parsing.
What it does
Use it to classify or score situations against a defined set of questions, including questions about image and video inputs.
- Availability
- Downloadable from Hugging Face; can be served with SGLang
- Pricing model
- Not established
- Deployment
- Tested on a single H200; SGLang serving instructions provided
- Privacy
- Not established
- Licence
- Apache-2.0
Capability
Modality
Access
Deployment
Pricing
Target user
Evidence
6 claims- Clef-Flash is a 9B multimodal model that returns a probability for every allowed option of every question in a single forward pass. [1]
probability for every allowed option of every question in a single forward pass.
Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights. - It accepts state as text, JSON, images, or video. [1]
It reads the state as text, JSON, images, or video, and returns a
Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights. - It does not produce free-form text or require output parsing. [1]
free-form text generation and no output parsing.
Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights. - The model files and usage code are available from the Cloudflare/clef-flash Hugging Face repository.
- Clef-Flash can be served with SGLang, with launch instructions for H200, B200, and B300. [1]
launch commands on H200, B200, and B300.
Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights. - The release is under the Apache-2.0 license. [1]
Released under the Apache-2.0 license, following the base model
Hugging Face. Quoted text is outside CC BY 4.0 and keeps its original rights.
Sources
1
1
Hugging FaceOwner-controlled
https://huggingface.co/Cloudflare/clef-flash
Retrieved 2026-10-28 00:15 UTC
Verified
Versions
3| Version | Effective | Change | Changed fields | State |
|---|---|---|---|---|
| 3 | 2026-10-10 12:11 UTC | material change | availability, deployment | Current |
| 2 | 2026-10-02 19:41 UTC | material change | access, availability | Current |
| 1 | 2026-10-01 20:36 UTC | publication | None | Current |
In briefs
2- Backburner leads 25 AI developments in week 40, 2026 Weekly brief 2026-W40
- Perplexity introduces a decision model fine-tuned from Qwen3.8-27B Daily brief 2026-10-01
Original catalog content is licensed under CC BY 4.0, attributed to wowsuchai. Source quotations and linked content keep their original rights.